1. Field of the Invention
The present invention relates to a field of camera calibration, more particularly to methods and devices for camera calibration from a single perspective image.
2. Description of Related Art
Camera calibration has always been an essential component of photogrammetric measurement and computer vision. Accurate camera calibration is necessary for extracting precise and reliable three-dimensional (3D) geometric information from two-dimensional (2D) plane images captured by a camera. Each 3D point in the 3D space may correspond to one 2D point in the 2D plane image captured by the camera. The relationship between the 3D points in the 3D space and the 2D points in the 2D plane image is represented as a perspective projection model of the camera. The perspective projection model comprises a plurality of intrinsic and extrinsic parameters of the camera. The intrinsic parameters are optical parameters of the camera, such as a focal length f, principal points and lens distortion parameters, etc. The extrinsic parameters are the pose and the position parameters of the camera, such as the tilt angle, the pan angle, the yaw angle and the height of the camera, etc. In camera calibration, the intrinsic parameters and the extrinsic parameters are estimated to construct the perspective projection model.
One of the conventional camera calibrations from a single perspective image is implemented based on a prearranged 3D control scene containing a plurality of known 3D control points. The camera captures a plane image of the 3D control scene. Points in the plane image corresponding to the known 3D control points are determined. The known 3D control points and the corresponding points in the plane image are substituted into the perspective projection model to calculate the unknown parameters including the intrinsic parameters and the extrinsic parameters.
The camera calibration depending on the 3D control scene has higher calibration accuracy. When the applications require high accuracy and the parameters of the camera are not frequently changed, the 3D control scene method is a preferable choice. However, implementing the 3D control scene method is a complex and difficult manual arrangement, and it is not always available for every applications.
Camera self-calibration is developed to calibrate the camera automatically and not depending on known reference objects. However, the current technology in the camera self-calibration is not very reliable and has lower calibration accuracy.
Thus, there is a need for improved techniques for accurate camera calibration from a single perspective image to overcome the above disadvantages.